SAS AI

Overview

SAS AI refers to the artificial intelligence and machine learning capabilities embedded within the broader SAS analytical software portfolio, with a particular focus on the SAS Viya platform. Established in 1976, SAS has developed a comprehensive ecosystem for data management, advanced analytics, and business intelligence, evolving to integrate modern AI techniques. The platform is designed for enterprise-grade deployments, offering tools for the entire analytical lifecycle: data access, preparation, exploration, model building, deployment, and monitoring.

SAS AI is positioned for organizations requiring precision, governance, and scalability in their analytical operations. Its strengths lie in complex statistical modeling, time-series analysis, and optimization, areas where SAS has historical depth. The platform supports various programming interfaces, including Python, R, Java, and its proprietary SAS language, allowing data scientists and developers to interact with its analytical engines. This flexibility aims to accommodate diverse skill sets within an enterprise data science team.

The primary use cases for SAS AI include predictive analytics in finance, fraud detection, risk management, customer intelligence, and operational optimization across sectors like healthcare, government, and manufacturing. For instance, in regulated industries such as banking and pharmaceuticals, the platform's auditing and explainability features are utilized to meet compliance requirements. Its hybrid cloud deployment options further extend its applicability, enabling organizations to run workloads on-premises, in private clouds, or on public cloud infrastructure like AWS, Azure, and Google Cloud, as detailed in SAS documentation for cloud deployment. DataRobot, a competitor in the automated machine learning space, also emphasizes hybrid cloud deployment options, reflecting a common enterprise requirement for flexibility across infrastructure types.

SAS Viya, a key component of the SAS AI offering, provides a cloud-native architecture designed for distributed processing and in-memory analytics. This architecture supports the execution of AI and machine learning models at scale, handling large datasets and complex computational tasks. The platform's integrated environment aims to streamline the transition from data exploration to model production, addressing challenges associated with deploying models into operational systems. The developer experience is supported by comprehensive APIs and SDKs, facilitating programmatic access and integration with existing enterprise applications, though deep domain knowledge of SAS architecture can be beneficial for complex integrations, as noted in developer resources.

Key features

• End-to-end AI/ML Lifecycle Management: Tools for data preparation, feature engineering, model development, validation, deployment, and monitoring within a unified platform.
• Advanced Statistical and Machine Learning Algorithms: Access to a library of algorithms covering traditional statistics, supervised and unsupervised learning, deep learning, natural language processing, and computer vision.
• Explainable AI (XAI): Capabilities to interpret model predictions and understand feature importance, supporting transparency and regulatory compliance.
• Model Governance and Risk Management: Features for model versioning, auditing, performance monitoring, and risk assessment, particularly relevant for regulated industries.
• Multi-language Support: Allows data scientists to work in Python, R, Java, Go, or the SAS language, integrating with existing codebases and skill sets.
• Hybrid Cloud Deployment: Supports flexible deployment across on-premises, private cloud, and public cloud environments (AWS, Azure, Google Cloud).
• Visual Analytics and Reporting: Integrated tools for data visualization, dashboard creation, and interactive reporting to communicate analytical insights.
• Automated Machine Learning (AutoML): Features to automate aspects of model selection, hyperparameter tuning, and feature engineering to accelerate model development.

Pricing

SAS offers custom enterprise pricing for its AI and analytics solutions, which typically involves discussions with their sales team to determine specific licensing models based on organizational needs, usage, and deployment scale. This approach is common among enterprise software providers, reflecting the complexity and customizability of their offerings.

For academic users, SAS provides SAS OnDemand for Academics, a free cloud-based offering that includes access to SAS software for teaching and learning purposes. For detailed enterprise pricing inquiries, organizations are directed to the official SAS pricing page to engage with their sales team.

Common integrations

• Cloud Platforms: Integration with major cloud providers including AWS, Microsoft Azure, and Google Cloud Platform for scalable deployments.
• Data Warehouses and Databases: Connectivity to various enterprise data sources such as Snowflake, Databricks, Oracle, SQL Server, and Hadoop.
• Programming Languages: SDKs and APIs for programmatic interaction using Python, R, Java, and Go, enabling integration with custom applications.
• Business Intelligence Tools: Integration capabilities with third-party BI solutions through data export and API access.
• ETL Tools: Connectors and functions for data extraction, transformation, and loading from diverse enterprise systems.

Alternatives

• Databricks: A data lakehouse platform offering unified data, analytics, and AI capabilities, built on Apache Spark.
• H2O.ai: Provides open-source and commercial AI platforms, including H2O-3 and H2O Driverless AI, focusing on automated machine learning.
• DataRobot: An automated machine learning platform designed to empower users with varying technical expertise to build and deploy AI models.
• IBM Watson Studio: A data science and machine learning platform on IBM Cloud, offering tools for data preparation, model building, training, and deployment.
• Google Cloud Vertex AI: A unified platform for machine learning development, offering MLOps tools and access to Google's AI infrastructure.

Getting started

To get started with SAS AI using Python, you can utilize the SAS Viya platform's REST APIs or the saspy package, which provides a Python interface to SAS. The following example demonstrates connecting to a SAS Viya environment and running a simple data step using saspy.

import saspy

# Configure saspy connection. This example assumes a basic configuration.
# For production, use secure methods for credentials and more robust connection details.
# Refer to the official saspy documentation for detailed configuration options:
# https://sassoftware.github.io/saspy/getting-started.html

sas = saspy.SASsession(cfgname='your_sas_config_name')

# Check if the connection is successful
if sas.sascfg.saspath is not None:
    print(f"Successfully connected to SAS at: {sas.sascfg.saspath}")
else:
    print("Failed to connect to SAS. Check your configuration.")
    exit()

# Create a simple SAS DataFrame (in-memory table in SAS)
data = {
    'ID': [1, 2, 3, 4, 5],
    'Value': [10, 20, 15, 25, 30]
}
sas_df = sas.df2sd(data, table='MyData', libref='WORK')

print("\nSAS DataFrame created:")
print(sas_df.head())

# Run a simple SAS DATA step to calculate the mean
sas_code = """
PROC MEANS DATA=WORK.MyData N MEAN STDDEV;
    VAR Value;
RUN;
"""

print("\nRunning SAS code:\n")
output = sas.submit(sas_code)
print(output['LST'])

# Disconnect from SAS session
sas.endsas()
print("\nDisconnected from SAS.")

This Python code snippet connects to a configured SAS session, transfers a Python dictionary into a SAS data frame, and then executes a PROC MEANS statement to perform basic descriptive statistics. To run this, you would need to have saspy installed (pip install saspy) and a valid SAS configuration file (sascfg.py) set up to point to your SAS environment. More advanced examples, including model building and deployment, are available in the SAS developer documentation and through the main SAS documentation portal.